[jakeway]

You can write code with pencil and paper

[jjp]

Yes but it's a little harder to prove that the algorithm you've described works.

[chrisseaton]

But almost nobody proves that programs they write on computers work either. When was the last time you proved a program correct?

[poteeweet]

They didn't mean "prove" as in mathematically "prove" than an algorithm works, they meant it in mechanical terms (does it compile, does it run, does it past test inputs, etc.).

[xixi77]

That is not a proof by the CS standard though -- well, unless the job is defined as "write a program that processes these N test inputs known in advance", in which case a program that prints the expected hardcoded output in response to each of the expected inputs and otherwise returns an error would be a good implementation :)

I personally avoid using a word "proof" to mean anything other than a mathematical proof. It can be done with code, for sure, but as grandparent said, usually isn't.

[robertbenjamin]

You don't prove that it's "correct," as there is no such thing as a "correct" program. You prove that it works.

[xixi77]

I think you may be assuming that the purpose of a CS degree is to teach people how to write programs that work, while it's really not :)

And there is such a thing as a correct implementation of an algorithm. Running it does not exactly prove things either, and it might not necessarily work on actual hardware, but it does show that the person understands at least part of the material.